Acta Materialia 213 (2021) 116971

Contents lists available at ScienceDirect

Acta Materialia
journal homepage: www.elsevier.com/locate/actamat

Full length article

Sustainable steel through hydrogen plasma reduction of iron ore:

Process, kinetics, microstructure, chemistry
I. R. Souza Filho a,∗, Y. Ma a, M. Kulse a, D. Ponge a, B. Gault a,b, H. Springer a,c, D. Raabe a
a
Max-Planck-Institut für Eisenforschung, Max-Planck-Str. 1, 40237 Düsseldorf, Germany
b
Department of Materials, Imperial College London, South Kensington, London SW7 2AZ, United Kingdom
c
Institut für Bildsame Formgebung, RWTH Aachen University, Intzestr. 10, 52072 Aachen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Iron- and steelmaking is the largest single industrial CO2 emitter, accounting for 6.5% of all CO2 emis-
Received 6 December 2020 sions on the planet. This fact challenges the current technologies to achieve carbon-lean steel production
Revised 16 March 2021
and to align with the requirement of a drastic reduction of 80% in all CO2 emissions by around 2050.
Accepted 3 May 2021
Thus, alternative reduction technologies have to be implemented for extracting iron from its ores. The
Available online 10 May 2021
hydrogen-based direct reduction has been explored as a sustainable route to mitigate CO2 emissions,
Keywords: where the reduction kinetics of the intermediate oxide product Fex O (wüstite) into iron is the rate-
Steel limiting step of the process. The total reaction has an endothermic net energy balance. Reduction based
Iron ore reduction on a hydrogen plasma may offer an attractive alternative. Here, we present a study about the reduction
Hydrogen plasma of hematite using hydrogen plasma. The evolution of both, chemical composition and phase transforma-
Atom probe tomography tions was investigated in several intermediate states. We found that hematite reduction kinetics depends
Microstructure
on the balance between the initial input mass and the arc power. For an optimized input mass-arc power
ratio, complete reduction was obtained within 15 min of exposure to the hydrogen plasma. In such a
process, the wüstite reduction is also the rate-limiting step towards complete reduction. Nonetheless,
the reduction reaction is exothermic, and its rates are comparable with those found in hydrogen-based
direct reduction. Micro- and nanoscale chemical and microstructure analysis revealed that the gangue
elements partition to the remaining oxide regions, probed by energy dispersive spectroscopy (EDS) and
atom probe tomography (APT). Si-enrichment was observed in the interdendritic fayalite domains, at the
wüstite/iron hetero-interfaces and in the oxide particles inside iron. With proceeding reduction, however,
such elements are gradually removed from the samples so that the final iron product is nearly free of
gangue-related impurities. Our findings provide microstructural and atomic-scale insights into the com-
position and phase transformations occurring during iron ore reduction by hydrogen plasma, propelling
better understanding of the underlying thermodynamics and kinetic barriers of this essential process.
© 2021 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction ing, a drastic reduction of 80% in all CO2 emissions is targeted

Annually, 2.6 billion tons of iron ore (mostly hematite) are con-
verted into steel by the integrated blast furnace (BF) and basic
oxygen furnace (BOF) route [1], accounting for approximately 70%
of the global steel production. The remaining 30% is realized by
melting steel scraps and directly reduced iron (the latter is also
referred to as sponge iron) in electric arc furnaces (EAF). On av-
erage, about 2.1 tons of CO2 are produced per ton of crude steel
[2]. This number corresponds to about 6.5% of all CO2 emissions
on the planet and makes iron- and steelmaking the largest in-
dustrial individual emitter of CO2 [3]. To mitigate global warm-
∗ tion conducted in shaft furnaces using reformed natural gas (e.g.
Corresponding author.
E-mail address: i.souza@mpie.de (I. R. Souza Filho).
until 2050 [4–6]. Thus, disruptive technology changes in iron ore
reduction must be urgently implemented, already within the next
years.
The use of hydrogen instead of carbon for iron ore reduction
is currently explored as an alternative sustainable route to miti-
gate the CO2 emissions [7–9]. The direct reduction of iron ore pel-
lets by pure molecular hydrogen above 570 °C occurs with the
intermediate formation of other iron oxide variants, viz. Fe2 O3
(hematite) → Fe3 O4 (magnetite) → Fex O (wüstite) → Fe (iron)
[7–22]. Although the net energy balance of the overall process is
endothermic [10], studies demonstrate that its reduction kinetics
can be reasonably faster than that of commercial direct reduc-
the Midrex process) [8,18]. This observation motivates the current

https://doi.org/10.1016/j.actamat.2021.116971
1359-6454/© 2021 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
I. R. Souza Filho, Y. Ma, M. Kulse et al.
global investments in hydrogen-based direct reduction pilot plants
[9].
Hydrogen plasma offers a viable alternative for carbon-neutral
iron making. Its high energy and enhanced density of H radicals
and exited states help to overcome the reaction’s activation barrier
[23–27] and has the potential for enhancing the Fex O reduction
rates by an order of magnitude [28–30], enabling iron conversion
to reach commercially viable rates. Also, hydrogen plasma-based
reduction allows the production of liquid iron in one single step,
in which the input fine ores are melted and reduced simultane-
ously without the need for intermediate agglomeration or refine-
ment processing, as the melting point of iron oxide (1565 °C) only
slightly exceeds that of iron (1538 °C) [31,32].
During the hydrogen plasma reduction (HPR), a plasma arc
zone is generated between an electrode and the input iron ore
(e.g. hematite) [33]. In this zone, the ore can be melted and re-
duced by hydrogen in both molecular and plasma states. The lat-
ter is composed of vibrationally ionized (H+ , H2 + and H3 + ), ex-
cited (H∗ ) and atomized (H) species, which are formed through
the mutual elastic and inelastic collisions of hydrogen particles
with electrons [34,35]. Under such conditions, the corresponding
degree of hydrogen dissociation is determined by the competing
ionization (e.g. e− + H2 → H2 + + 2 e− ) and recombination (e.g.
H+ + e− → H) events [34]. The high energy carried by the hy-
drogen plasma is partially released at the reaction interface, gen-
erating a large amount of local heat [36,37]. This heat diminishes
the need for external power supply and promotes high power-
efficiency to the process [32]. Thus, the thermo-kinetics advantages
observed in HPR depend directly on the concentration of hydro-
gen plasma radicals that are able to reach the reaction interface
[25,34,36,37].
Plasma discharges are generated by electric arcs in conven-
tional EAFs to melt steel scraps. Thus, it is conceivable that es-
tablished industrial electric furnaces can be modified to be used
for HPR purposes, with only small (e.g. 5–10%) H2 partial pres-
sure. Due to the limited number of scientific investigations on this
topic [25,28,29,34–39], closer analysis of the underlying reaction
steps and of the influence of gangue-related tramp elements at the
nano-scale may help to identify conditions for developing HPR pro-
cesses with high kinetics and efficiency.
Here, we performed interrupted Ar-10%H2 -plasma reduction
experiments of hematite to study and evaluate the HPR pro-
cess stepwise in high detail. Emphasis is placed on the evolving
nano-chemistry, interface structure and composition and the phase
transformation kinetics. We observe that the hematite reduction
kinetics depends on the balance between input material mass and
arc power. Wüstite reduction is the rate-limiting step of the pro-
cess and the gangue elements are gradually removed from the
samples with an increase in the reduction cycles. This seems to
be particularly important for the case of Si, which in conventional
H-based direct reduction can decorate and impede the hetero-
interface between Fe and the Fex O. We show that low-impurity-
content iron can be produced by hydrogen plasma with reduction
rates comparable to those found in the direct reduction conducted
using molecular hydrogen. Our findings advance the understanding
of the succession of phase transformations and chemical evolution
during HPR and their influence on the thermodynamic and kinetic
barriers to scale up green iron production.
2. Experimental procedures
2.1. Input material and hydrogen plasma-based reduction
Commercial hematite pieces were used in this study and their
chemical composition is presented in Table 1. The density of the
hematite pieces was measured as 5.2983 ± 0.0016 g/cm3 via ar-
2
Acta Materialia 213 (2021) 116971
gon pycnometer analysis. Samples of hematite with an average
weight of 9 g were introduced into an arc-melting furnace, which
was equipped with a tungsten electrode (diameter of 6.34 mm)
and filled with a gas mixture of Ar-10%H2 (total pressure of 900
mbar, chamber volume of 18 L). To melt and reduce the samples,
a plasma arc was ignited between the tip of the electrode and the
input material at a voltage of 44 V and a current of 800 A. The
average distance between the electrode and the hematite samples
was 6 mm. Fig. 1(a) shows a schematic representation of the ap-
paratus.
To track the progress of both, the phase transformations and
chemical composition over the course of the reduction process, the
samples were submitted to interrupted reaction cycles. To obtain
specimens in different partially reduced states, the hematite pieces
were subjected to sequential melting, solidification and remelting
series at a pre-defined number of times (viz. 1, 2, 5, 10, 15 and 30
times) with a corresponding exposure time to the hydrogen plasma
at 1 min for each cycle. During the 1-min cycles, complete melting
of the samples was achieved with an average exposure time of 6 s.
Once the plasma arc was switched off, rapid solidification condi-
tions were achieved by means of the water-cooled copper hearth
on which the hematite pieces were molten and reduced. To ensure
the proper stoichiometry conditions (i.e. a reductant atmosphere)
and maintain the progressing reduction, the furnace chamber was
replenished by a fresh mixture of Ar-10%H2 after the completion
of every individual melting and solidification cycle. In summary,
one single process cycle comprises the following steps: 1) charging
of the furnace with the oxide sample and flooding of the cham-
ber with fresh Ar-10%H2 gas mixture; 2) simultaneous plasma arc
melting and reduction of the sample; 3) intermediate interruption
of the plasma arc operation and solidification of the (partially or
fully reduced) sample, and 4) replenishing of the furnace’s cham-
ber with fresh Ar-10%H2 gas mixture for the continuation of the
process (i.e. succeeding cycle). The sequential steps of this proce-
dure are summarized in Fig. 1(b). Results will be reported as a
function of the total time of exposure to the plasma experienced
by each sample. The reduced samples (Fig. 1c) were further char-
acterized via microstructural, chemical and phase transformation
analysis, in part down to atomic scale.
2.2. Microstructural characterization
The solidified samples were cut, embedded and metallograph-
ically prepared for high-resolution scanning electron microscopy
analysis, including electron backscatter diffraction (EBSD) and elec-
tron channeling contrast imaging (ECCI). EBSD maps were acquired
using a JEOL-6500F field-emission gun microscope. The mapped ar-
eas were further imaged with the aid of the ECCI technique us-
ing a Zeiss Merlin scanning electron microscope (SEM). Energy-
dispersive X-ray spectroscopy (EDS) analyzes were also conducted
in representative areas of the samples, using a Zeiss Merlin SEM
and an accelerating voltage of 15 kV.
2.4. Chemical composition analysis
To obtain the average weight percentage of elements at each
stage of reduction, chemical composition analysis was conducted
for the solidified samples containing both, the iron and untrans-
formed oxide portions (Fig. 1c). The content of metals was de-
termined using inductively coupled plasma optical emission spec-
trometry (ICP-OES). The oxygen content was determined via reduc-
tion fusion under a helium atmosphere and subsequent infrared
absorption spectroscopy. Carbon and sulfur were analyzed via com-
bustion followed by infrared absorption spectroscopy. Iron nuggets
extracted from the partially reduced samples (5, 10, 15 min) were
also analyzed.
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Table 1
Chemical composition of the initial hematite measured by inductively coupled plasma opti-
cal emission and infrared absorption spectroscopies (in wt.%).

Fe Si Al Na Mn Cu Ca Mg Ti

70.228 0.104 0.077 0.017 0.0094 0.005 0.004 0.0024 0.0019

W C P S O

0.002 0.006 0.0103 0.133 29.4

Fig. 1. (a) Schematic illustration of the arc melting furnace equipped with a tungsten electrode and charged with Ar-10%H2 gas mixture. The input hematite is placed on
the water-cooled copper hearth inside the furnace before the reduction process. (b) Sequential steps for the reduction process. (c) Hematite pieces reduced with hydrogen
plasma under different plasma exposure times: 1 min; 2 min; 5 min; 10 min; 15 min and 30 min. The red arrows indicate the presence of millimetric-scale transformed
iron in the bottom of the samples. The upper part of the samples corresponds to the untransformed remaining oxide portion.

2.5. Thermodynamic simulations 2.6. Atom probe tomography

Based on the average chemical composition of representative
samples, the corresponding metastability phase calculations, in
which the face-centered cubic iron was excluded, were conducted
using the software ThermoCalc, coupled with the metal oxide so-
lutions TCOX10 database. For comparison, the corresponding phase
equilibrium diagrams were also calculated and reported in the
Supplementary Material (Fig. S.1).
To characterize the atomic-scale elemental distribution across
the wüstite-iron interface, atom probe tomography (APT) was em-
ployed. The APT specimens were prepared using the site-specific
lift-out method at the wüstite-iron interface (Fig. S.2) of the 5 min-
reduced sample by a FEI Helios NanoLab600i dual-beam focused
ion beam/scanning electron microscopy instrument. The APT mea-
surements were conducted using a Local Electrode Atom Probe

3
I. R. Souza Filho, Y. Ma, M. Kulse et al.
(LEAP, Cameca Instruments Inc.) 50 0 0 XS with a detection effi-
ciency of ~ 80%. The instrument was operated in laser-pulsing
mode at a wavelength of 355 nm, laser energy of 40 pJ, and laser
pulse frequency of 200 kHz. The base temperature in the analysis
chamber was maintained at 50 K during the measurements. The
reconstruction of the three-dimensional atom maps and data anal-
ysis was carried out using the commercial software AP Suite 6.0.
2.7. Phase quantification analysis
The partially reduced samples were hammered to separate the
formed iron portion, i.e. nuggets, from the remaining oxides. The
obtained iron and oxide portions were weighted in a precision
analytical balance (accuracy of ± 0.0 0 01 g). The remaining oxide
was further ground and probed by X-ray diffraction measurements
(XRD) using a diffractometer D8 Advance A25-X1, equipped with
a cobalt Kα X-ray source, operated at 35 kV, 40 mA. The diffrac-
tion data were processed using the software Brucker Topas v. 5.0.
Rietveld refinement was employed to quantify the weight fraction
of the constituents in the powder. To determine the iron conver-
sion kinetics, the weight of the iron within the oxide portion was
summed with that of the iron nugget extracted from the same
sample.
3. Results
3.1. Macrostructural analysis
Fig. 1(c) shows the cross-section of the samples reduced with
different exposure times to the hydrogen plasma. After 1 and
2 min of reduction, no visual evidence for iron is observed. From
5 to 15 min, millimeter-sized iron portions (indicated by the red
arrows in Fig. 1c) appear brighter than the remaining oxides, the
latter mostly located in the top part of the samples. Total conver-
sion is reached after 30 min. Also, the average porosity of approx-
imately 35% is found in the partially reduced samples (from 1 to
15 min). Details regarding the porosity analysis are given in the
Supplementary Material (Fig. S.3). The completely reduced sample
is free of macro-pores.
3.2. Microstructural analysis
Fig. 2 shows the microstructural analysis of the specimen par-
tially reduced for 1 min. The analyzed area is indicated in Fig. 2(a)
by the red frame. Fig. 2(b) depicts the corresponding ECCI-image
and reveals a columnar grain structure with hundreds of μm in
length. The EBSD inverse pole figure (IPF) map in Fig. 2(c) shows
that the grains grow with a strong crystallographic texture with
the majority of the 001 directions aligned perpendicular to the
sample’s surface (i.e. parallel to the solidification direction). Such
a columnar microstructure is consistent with that observed also in
other unidirectional solidification processes, governed by fast heat
extraction, imposed here by the water-cooled hearth. The phase
map displayed in Fig. 2(d) shows the coexistence of magnetite
and wüstite in this sample, represented in purple and green, re-
spectively. The remaining magnetite, when found dispersed within
wüstite, shows a cuboidal morphology, Figs. 2 (e-g).
Fig. 3(a) is a similar overview image of the sample partially re-
duced for 2 min. The analyzed area is indicated by the red frame
in this figure. Fig. 3(b) shows that small iron volumes with sizes of
up to 10 μm are readily distinguished due to their bright contrast
(arrowed). Fig. 3(c) shows details of a drop-shape iron domain sur-
rounded by smaller satellite iron islands. The same area was fur-
ther probed via EBSD and the corresponding maps are displayed
in Figs. 3 (d-h). Both, the EBSD phase (Fig. 3f) and oxygen distribu-
tion (Fig. 3g) maps confirm the presence of iron in the microstruc-
4
Acta Materialia 213 (2021) 116971
ture. Yet, they evidence that small oxide particles are found within
iron (Fig. 3f and g) as well as the wüstite/ferrite phase interfaces
are enriched in silicon (Fig. 3h). For all investigated areas of this
sample, wüstite was the remaining iron oxide phase.
Fig. 4 shows the microstructural analysis of the sample par-
tially reduced for 5 min. EBSD maps, shown in Fig. 4(b), were ac-
quired for the region indicated by the red frame in Fig. 4(a). The
oxygen distribution map in Fig. 4(c) evidences small iron domains
(arrowed) dispersed within wüstite. Also, Fig. 4(c) reveals a Si en-
richment at the wüstite/iron phase interfaces as well as in the in-
terdendritic regions (backscatter electron image, BSE). The small
volume fraction of this constituent hindered its quantification via
XRD. However, similar analysis conducted for the sample partially
reduced for 10 min (Supplementary Material, Fig. S.6) suggests that
it is fayalite (Fe2 SiO4 ).
Oxide particles are observed within iron, as revealed by the sec-
ondary electron (SE) image in Fig. 4(d). Their chemical composition
can be qualitatively evaluated via the corresponding EDS maps.
Fig. 4(d) shows a large particle (> 1 μm) that has a Si-enriched
core surrounded by a Fe-O-bearing shell. The smaller particles are
composed of iron and oxygen. The formation mechanisms of these
oxide particles can be associated with the reduction kinetics and
with hydrodynamic aspects of the molten bath flow. In terms of
kinetics, one can assume that the molten oxide, when in direct
contact with the plasma, does not in some cases get entirely re-
duced. In this scenario, small oxygen-enriched molten pools re-
main partially reduced within the volume majorly composed of
liquid iron. When considering that the interaction between the arc
plasma and the molten material is a rather complex phenomenon,
one might also expect that turbulences of the melt flow would al-
low partially reduced domains (e.g. Fex O) to get trapped inside the
liquid iron, due to bath dynamics. The fast solidification rate im-
posed by the water-cooled copper hearth might be a reason that
these molten oxygen-enriched (i.e. partially reduced) pools solidify
in the form of oxide particles. In this case, their chemistry should
not vary drastically from that of the larger oxide portions located
in the upper part of the sample (composed of wüstite and fayalite).
Thus, we could plausibly expect compositions ranging from Fex O to
Fe2 SiO4 for these particles as well.
Fig. 5 shows the atom probe tomography analysis of a wüstite-
iron interface in the sample partially reduced for 5 min. The cu-
mulative field evaporation histogram is displayed in Fig. 5(a), and
the red arrows highlight the wüstite-iron interface. The reduced
iron reveals the crystallographic poles in the field evaporation his-
togram. The crystallographic pole marked by the red dot is iden-
tified as (011). The corresponding {011} atomic planes are visu-
alized in Fig. 5(b) and the interplanar spacing is approximately
0.20 nm. This value matches well with the interplane spacing of
body-centered cubic iron. Such crystallographic information is also
helpful for the improvement of the spatial accuracy of the recon-
struction of the 3D atom maps [40,41]. Fig. 5(c) shows the distribu-
tion of iron and oxygen atoms in the APT specimen. The interface
of the wüstite and iron is marked by a 20 at.% oxygen isocompo-
sition surface. The reduced iron contains more than 98 at.% iron,
as shown by the composition profile calculated across the inter-
face in Fig. 5(d) and summarized in Table 2. In contrast, the APT
analysis reveals that wüstite is mainly composed of ~ 59 at.% iron
and 40 at.% oxygen, yielding a stoichiometric ratio ~ 3:2, which
differs from the theoretically expected stoichiometric ratio ~ 1:1.
The lower oxygen concentration measured by APT is most likely
due to the loss of neutral oxygen during field evaporation as des-
orbed neutral oxygen atoms and also O2 molecules, which cannot
be measured by the detector [42–44]. The contents of trace ele-
ments (e.g. Al, Ca, Mg, Si, Co, Ni, and V) are considerably lower in
the reduced iron compared to those within the oxide, as shown in
Table 2. In addition, a sharp enrichment of silicon is observed at
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 2. Microstructural characterization of the sample partially reduced for 1 min. (a) Overview of the sample. The red frame shows the chosen area for the correlative EBSD-
ECCI probing approach. (b) ECCI-image of the area delimited by the red frame in (a). (c) EBSD inverse pole figure (IPF) of the region highlighted by the dashed frame in (b).
The crystallographic orientations shown in this map are perpendicular to the normal direction of the sample, i.e. parallel to the solidification direction; (d) corresponding
EBSD phase map showing magnetite and wüstite in purple and green, respectively. (e) Enlarged view of the area delimited by the frame in (d); (f) corresponding ECCI-image
of the area displayed in (e). (g) Enlarged view of the area delimited by the red frame in (f), revealing the presence of cuboidal-shaped remaining magnetite dispersed in a
columnar-shaped wüstite grain.

Table 2
Chemical composition of reduced bcc-iron and wüstite of the sample reduced for 5 min analyzed by atom probe tomography (in at.%).

Fe Al Ca Mg Mn O Si Co Ni V

bcc-iron 98.082 0.006 0.0003 0.004 0.035 0.742 0.013 0.005 0.003 0.004
± 0.302 ± 0.002 ± 0.0006 ± 0.002 ± 0.006 ± 0.175 ± 0.009 ± 0.002 ± 0.002 ± 0.002
wüstite 59.351 0.083 0.0016 0.044 0.060 39.89 3 0.063 0.035 0.007 0.071
± 0.219 ± 0.009 ± 0.0011 ± 0.006 ± 0.007 ± 0.206 ± 0.007 ± 0.006 ± 0.002 ± 0.007

the wüstite/bcc-iron interface, as demonstrated by the concentra-
tion profile across the interface in Fig. 5(e).
Fig. 6 shows the microstructural evaluation of the sample com-
pletely reduced into iron (30 min). Fig. 6(b) shows an ECCI-image
which reveals columnar grains subdivided by sub-grains. The mi-
crostructure is nearly devoid of oxide particles. The enlarged view
provided by Fig. 6(c) reveals a few worm-like inclusions, whose
volume percentage is only 0.10 ± 0.05%. A representative SE-EDS
analysis for one of these inclusions (Figs. 6d-g) suggests that they
are enriched in Si and O. The point EDS analysis performed within
it (yellow star in Fig. 6h) also reveals the presence of several
gangue elements, including Ca, Si, S and Cu, Fig. 6(i) [45].
3.2.1. Oxide particles within iron
The evolution of both, the size and fraction of oxide particles
within iron accompanies the reduction progress, as demonstrated
by Fig. 7(a). This figure shows that the particle percentage reaches
up to 3.3 ± 0.6% with 15 min of reduction, but sharply drops to
nearly zero after 30 min (complete reduction). The weighted av-
erage size of the particles evolves up to 0.95 μm with 15 min,
but it is considerably reduced to 0.16 μm in the final iron product.
Figs. 7(b) and (c) compare the microstructures of the iron portions
obtained with 15 and 30 min of reduction, respectively. The final
stages of the process (i.e. from 15 min onwards) seem to enable
the deoxidation of iron.
3.3. Chemical analysis
Fig. 8 shows the changes in the chemical composition of the
samples during reduction. Here, it is important to clarify that these
values represent the chemical composition of the entire specimen
without distinguishing the iron and untransformed oxide portions.
For the sake of comparison, the chemical composition of the initial
hematite is also displayed in this figure (viz. at 0 min). Fig. 8(a)
shows that the oxygen content of the samples gradually decreases
until reaching 0.0064 wt.% (64 ppm) in the final iron product.
An interesting general trend is that the quantities of Al, Mn
and Si (Fig. 8b), P and S (Fig. 8d) are remarkably reduced over the
course of the reduction. This purification process can be achieved

5
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 3. Microstructural characterization of the sample partially reduced for 2 min. (a) Overview of the sample. The red frame shows the chosen area for the correlative
EBSD-ECCI probing approach. (b) ECCI-image of the area delimited by the red frame in (a). The white arrows evidence the presence of small domains of iron. (c) ECCI-image
showing an enlarged view of the region delimitated by the yellow frame in (a). The corresponding EBSD maps of the area displayed in (c) are as follows: (d) image quality
(IQ) map; (e) inverse pole figure (IPF) map in which the crystallographic orientations are shown perpendicular to the normal direction of the sample, i.e. parallel to the
solidification direction; (f) phase map where wüstite and ferrite are represented by green and red respectively; (g) oxygen distribution and (h) silicon distribution maps. The
corresponding captions are properly indicated in the bottom left-hand side of the figure.

Table 3
Chemical composition evolution of the produced iron measured by inductively coupled plasma optical emission and infrared absorp-
tion spectroscopies (in wt.%).

Fe Al C Ca Cu Mg Mn O P S Si W

5 min bal. 0.008 0.005 0.004 0.019 0.003 0.009 9.590 <0.002 <0.002 0.009 0.055
10 min bal. <0.001 0.005 <0.001 0.014 <0.001 <0.001 0.500 <0.002 <0.002 0.002 0.002
15 min bal. <0.001 0.005 <0.001 0.011 <0.001 <0.001 0.500 <0.002 <0.002 0.002 0.041
30 min bal. <0.001 0.006 <0.001 0.002 <0.001 <0.001 0.006 <0.002 <0.002 0.003 0.128

by sputtering induced by the plasma arc and/or by removal via
gas phases [32]. Although C, Ca and Mg concentrations slightly in-
crease up to 15 min of reduction, the final stages of the reduction
drop the corresponding Ca and Mg values to 10 ppm (Fig. 8c). The
C content in the final iron was found to be slightly higher than
that in the initial hematite.
Cu and W enrichment is observed up to 2 min of reduction as
a likely consequence of the melting of small parts of the hearth
and electrode (Fig. 1a), respectively (Fig. 8e). For longer times, the
Cu content is reduced to 20 ppm, which is lower than that found
in hematite (50 ppm). W values in turn vary with a non-defined
trend.
The chemical composition evolution of the iron portion formed
from 5 min of reduction onwards is given in Table 3. This result
shows that the produced iron is nearly free of harmful elements
such as P and S (< 20 ppm). Yet, the gradual deoxidation is es-
sential to the removal of oxide particles, as previously observed in
Fig. 7. The gangue elements are residual and, according to Fig. 6,
they are confined in the negligible fraction of the worm-like inclu-
sions.
3.4. Thermodynamic simulations
Based on the chemical composition of the samples partially re-
duced for 5, 10 and 15 min (Fig. 8), the corresponding metastability
of the phases were calculated as displayed in Figs. 9 (a-c), respec-
tively. These diagrams enable us to infer the likely metastable so-
lidification path of the samples. Due to the low contents of P and
S (Fig. 8), these elements were not considered in the calculations.

6
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 4. Microstructural characterization of the sample partially reduced for 5 min. (a) Overview of the sample. The red frame shows the area evaluated via EBSD whose
maps are displayed in (b). (c) Backscatter diffraction electrons (BSE) image of the frame displayed in (b). The oxygen and silicon distribution maps are also shown in (c). (d)
Secondary electrons (SE) image of the interior of iron, revealing oxide particles. The corresponding iron, oxygen and silicon distribution maps are also shown in this image.

7
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 5. Atom probe tomography (APT) analysis of the wüstite-iron interface in the partially reduced sample for 5 min. (a) Cumulative field evaporation histogram with the
(011) pole of the reduced iron (the arrows indicate the position of the interface); (b) spatial distribution map analysis showing the {011} lattice plane of the reduced iron; (c)
three-dimensional atom probe tomography maps of iron and oxygen, and the wüstite-iron interface is marked by a 20 at.% oxygen isoconcentration surface; concentration
profiles of (d) iron and oxygen, as well as (e) silicon relative to the position of the wüstite-iron interface.

W and Cu are incorporated as tramp elements during the melting
and thus were also neglected.
Fig. 9 shows that, under all reduction conditions probed in this
study, the Fe-enriched liquid starts solidifying into bcc-Fe (body-
centered cubic iron) at approximately 1530 °C, whereas the coexis-
tent oxygen-enriched liquid portion is replaced by wüstite at tem-
peratures below 1370 °C. At the final stages of the solidification,
the remaining liquid is transformed into fayalite, as indeed ob-
served in the interdendritic regions inside wüstite (Fig. 4 and Fig.
S.4). Figs. 9(b) and (c) also show that fayalite can decompose, via a
solid-state transformation, into two variants (viz., one enriched in
iron and the other doped with Ca) at temperatures around 10 0 0 °C.
The weight fractions of bcc-Fe and wüstite displayed in Fig. 9 are
in excellent agreement with the ones experimentally determined
and reported in Fig. 10.
For comparison, the corresponding equilibrium phase diagrams
are shown in Fig. S.1 of the Supplementary Material. Fig. S.1 shows
that the solid-state bcc-iron → fcc-iron transformation is expected
to occur between 910 and 1390 °C. However, due to the very
fast solidification conditions imposed by the water-cooled copper
hearth, such transformation is kinetically unfavored to proceed in
practice.
3.5. Phase transformations
The phase transformations as a function of the plasma expo-
sure time are shown in Fig. 10. In this figure, only the major
Fe/O-containing phases are considered, i.e. the very small fraction
of the interdendritic fayalite (0.73 wt.%, Supplementary Material)
within Fex O was neglected. Fig. 10(a) shows that hematite is re-
duced into a dual-phase magnetite (0.24) and wüstite (0.76) mix-
ture after 1 min of reduction, confirming the previous observations
made via microstructural analysis (Fig. 2). With 2 min of reduc-
tion, most of the sample is wüstitic (0.98) and the metallic iron
drops observed for this reduction condition (Fig. 3) represent a
weight fraction of 0.02. For longer reduction times, wüstite is grad-
ually transformed into iron, whose formation kinetics follows a sig-
moidal trend, Fig. 10(b). The complete reduction is achieved after
30 min and the corresponding iron conversion rate reaches a max-
imum (8 wt.%/min) at approximately 3.5 min.
4. Discussion
4.1. Iron conversion kinetics
Fig. 10(a) showed that the last 15 min of reduction via hy-
drogen plasma are accompanied by relative sluggish kinetics. In
other words, 75% reduction is achieved within the initial period of
15 min, whereas the last 25% also requires 15 min. This observation
indicates that the intermediate wüstite reduction imposes thermo-
dynamic and kinetic limiting barriers to the total conversion into
iron. In fact, the reduction of molten wüstite by hydrogen plasma
is the most unfavorable reaction when compared with the preced-
ing reduction steps, at temperatures ranging from 1600 to 20 0 0 °C
[34,36]. For example, the equilibrium Gibbs free energy changes
(G) for reducing wüstite, magnetite and hematite using H+ ions
at 1600 °C are approximately −2.910−3 kJ/mol, −3.910−3 kJ/mol

8
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 6. Microstructural characterization of the sample completely reduced after 30 min of exposure to hydrogen plasma. (a) Overview of the sample. (b) Corresponding
microstructure observed with the aid of the ECCI technique. (c) Enlarged view of the region delimitated by the yellow frame in (b). (d) Enlarged view of the area delimited
by the yellow frame displayed in (c), showing details of a worm-like inclusion. Corresponding elemental maps for (e) iron; (f) oxygen and (g) silicon. (h) Morphological
details of the inclusion displayed in (d). The yellow star represents the area probed via EDS. The obtained elemental spectrum is shown in (i).

9
I. R. Souza Filho, Y. Ma, M. Kulse et al.
Fig. 7. (a) Evolution of the volume fraction and average size of the oxide particles pinned inside iron. (b) ECCI-image of the microstructure of the iron formed in the 15-min
partially reduced sample (c) ECCI-image of the sample completely reduced after 30 min of exposure to the hydrogen plasma.
and −4.410−3 kJ/mol, respectively [34]. These energy values are
generally calculated assuming that all reductant species in the arc
plasma are H+ ions in thermodynamic equilibrium, i.e. neglect-
ing all the plasma particle collisions. In contrast to this idealized
exothermic plasma scenario, one must bear in mind that the solid
state wüstite reduction using molecular hydrogen (i.e. direct reduc-
tion) displays positive G values and, therefore, this net process
is thermodynamically not exothermic [34,36,46]. This means that
additional energy must be provided to drive the underlying redox
reaction. This endothermic energy balance of the oxide reduction
in the solid / gas mixture advocates the use of a hydrogen plasma
instead of regular molecular hydrogen, as the energy provided to
create a plasma produces a high density of very reactive radicals
that overcome the thermodynamic barriers readily and equips the
overall wüstite reduction with high efficiency.
Fig. 11 shows the diagram ‘microstructure versus oxygen con-
tent’ for the reduced samples. This diagram helps to observe how
the total oxygen content in a sample is partitioned into different
phases. The very small oxygen fraction found in both, the residual
fayalite and oxide particles within iron were not considered in the
results displayed in this figure. Based on the data shown in Fig. 11,
we fairly estimated the oxygen removal after each reduction stage
and the obtained results are displayed in Table 4. The amount of
hydrogen not consumed during the reaction was also estimated
through the calculation of mass balance, taking into account the
oxygen content in the samples (neglecting the small material loss
induced by arc sputtering), their respective phase fractions and the
hydrogen provided in the furnace chamber after each cycle, as also
depicted in Table 4.
10
Acta Materialia 213 (2021) 116971
It was proposed [28] that reduction reactions in HPR occur
through the following steps: (1) mass transfer of hydrogen and
oxygen respectively from the gas and molten oxide to the reaction
interface; (2) adsorption of hydrogen and adsorption/dissociation
of Fex O at the reaction interface; (3) chemical reaction; (4) desorp-
tion of H2 O from the interface; (5) mass transfer of H2 O away from
the interface. Ref. [28] also reveals that there is a critical value of
gas flow rate above which the mechanism of gas transfer is not the
rate-limiting step of the process (i.e. the reduction rate becomes
independent on the gas flow rate). If the furnace’s atmosphere
contains a stoichiometrically sufficient amount of reductant agents,
the overall reduction rate is determined by the slowest step which
is assumed to be step (3), i.e. the chemical reaction. Differently
from Ref. [28], our experiments were not conducted under con-
stant gas flux. However, by replenishing the furnace chamber with
fresh Ar-10%H2 gas mixture after the completion of one melting
cycle (Section 2.1 and Fig. 1b), we ensure sufficient stoichiometry
conditions for the progressing reaction, since the provided amount
of hydrogen is much higher than that consumed during each re-
action step (Table 4). Yet, according to the Baur-Glaessner diagram
modified to consider the atomic hydrogen existing at high temper-
atures in the plasma medium, iron is stable in the presence of at
least 60% of H2 molecules contained in a mixture of H2 O + H2
(viz. product water and excess of hydrogen), between 1600 and
2300 °C [32]. The corresponding ‘H2 in (H2 O + H2 )’ molar fractions
at each reduction time are also given in Table 4 and evidence that
the furnace atmosphere was always kept reducing, thus display-
ing favorable conditions for iron formation. Therefore, appropriate
stoichiometry is provided, and the plasma generation conditions
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Table 4
Oxygen removal during reduction and chemical composition of the furnace atmosphere. Total amount of hydrogen pro-
vided in the furnace chamber prior to reduction. Excess of hydrogen after a certain exposure time to the plasma. Corre-
sponding fraction of H2 in the mixture of H2 + H2 O after reduction.

Exposure time(min) Oxygen removal (%) Total provided H2 (mol) Excess H2 (mol) H2 in (H2 + H2 O)

1 16.7 0.065 0.041 0.63

2 21.1 0.131 0.099 0.76
5 35.7 0.327 0.271 0.83
10 60.5 0.654 0.551 0.84
15 78.9 0.981 0.836 0.85
30 ~100 1.962 1.558 0.79

are identical (44 V, 800 A). Due to these well-controlled chemical

Fig. 8. Chemical composition evolution during reduction. Changes in weight per-
centage in terms of (a) Fe and O; (b) Al, Mn, and Si; (c) C, Ca, and Mg; (d) P and S;
(e) Cu and W, where the logarithmic y-axis serves to facilitate the visualization.
boundary conditions, we can plausibly infer that the mass transfer
from the gas to the reaction interface has been similar during all
melting cycles, providing similar conditions to the reaction kinetics
[28,29].
The relative sluggish kinetics of wüstite reduction from 15 min
onwards is associated with its ever-lower fractions, i.e. the lower
the content of wüstite in the sample the more difficult is its re-
duction. This fact is most probably linked to physical characteris-
tics of the plasma arc process: the reduced iron has higher mass
density than the remaining wüstite. Consequently, iron tends to
sink to the bottom of the melt, leaving remaining wüstite on top
(similar to a slag) which gets thus exposed to the arc plasma (as
reduction can only occur on the melt surface). However, the den-
sity difference is not so significant that it can sufficiently counter-
act the dynamic material flow occurring in the melt. Such a dy-
namic flow is caused by the forces (magnetic and kinetic) imposed
by the arc, together with the localized heating on top (induced by
the contact point of the arc) and cooling on bottom of the melt
(by the water-cooled hearth). While this fact is favorable from a
point of desired homogeneity of the melt, it also constantly moves
the unreduced wüstite through the desired reaction zone directly
under the plasma arc, which, with decreasing wüstite fraction, be-
comes less and less probable. Furthermore, as the wüstite fraction
and oxide particle size decreases, the arc plasma preferably con-
tacts already reduced Fe, which exhibits less resistance to the elec-
tron transfer required [34,35,38,39]. These factors strongly depend
on the respective experimental conditions and can be addressed
by technical improvements (especially in an industrial EAF, which
typically uses lower temperature arc plasmas and lower arc power
to melt volumes) such as pulsed arcs, external magnetic fields or
deliberate pauses between arc cycles to induce oxide floatation. In
light of these observations, the most probable rate-limiting barriers
after 75% of reduction (Fig. 10a) can be considered to be the oxy-
gen mass transportation to the reaction interface and the reduction
reaction itself.
The reduction kinetics of molten oxides are dependent on sev-
eral experimental factors such as the electrode’s material and di-
ameter, length of the arc, ratio of the input mass to the arc power
and cooling. To provide better conditions for wüstite floatation
and ensure its exposure to the reaction zone, we increased the
input material weight to approximately 15 g, while all the other
parameters were kept constant. This second reduction procedure
was conducted as described in Section 2.1 and the obtained phase
evolution is displayed in Fig. 12(a). This figure shows that al-
ready 1 min of plasma exposure promotes total transformation of
hematite into wüstite. For longer times, the iron conversion follows
a sigmoidal trend, and total reduction is achieved after 15 min.
The maximum transformation rate (12 wt.%/min) occurs at ap-
proximately 7 min (Fig. 12b). This result means that the reduction
of the 15 g-hematite pieces proceeds two times faster than that
of the 9 g pieces, yet, at rates of about one order of magnitude
higher.

11
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 9. Metastability calculations for the samples partially reduced for (a) 5 min, (b) 10 min and (c) 15 min, conducted using the software ThermoCalc with the aid of the
metal oxide solutions database TCOX10. For the construction of these diagrams, austenite (fcc-iron) was excluded from the calculations.

Fig. 10. Phase transformation kinetics during hydrogen plasma-based reduction of the 9 g-hematite pieces. (a) Changes in the phase weight fractions of magnetite, wüstite,
and ferrite. (b) Iron conversion kinetics and corresponding transformation rates.

This second set of experiments confirms that a substantial ki-
netics enhancement can be achieved by adjusting the input mate-
rial mass, the mass-to-surface ratio, and consequently the molten
bath height, in relation to the plasma arc power and general fur-
nace conditions (bath dimensions, cooling power, etc.). It appears
that a larger molten bath facilitates mass transport of oxygen to
the reaction interface (i.e. promoting wüstite flotation), thus speed-
ing up the reaction. In this context, one must also consider that
the hydrodynamic features developed during plasma arc melt-
ing are rather complex and several other parameters such as gas
shear stress, Marangoni and electromagnetic forces, buoyancy, and
thermo-capillarity convection contribute to mass and heat distri-
bution throughout the liquid [31,38,47–52]. For the second set of
reduced samples, 1 min of exposure to the hydrogen plasma en-
ables the complete reduction of hematite into wüstite whereas ad-
ditional 14 min is required to reach pure iron. This result confirms
that the wüstite reduction is the rate-limiting step in the hydro-
gen plasma-based reduction of iron ores, even under conditions of
optimized kinetics.

12
I. R. Souza Filho, Y. Ma, M. Kulse et al.
Fig. 11. Diagram ‘microstructure versus oxygen content’ for the 9 g-hematite pieces
reduced via hydrogen plasma under different exposure times.
4.2. Comparison with the kinetics data reported in the literature
Seftejani et al. [35] investigated the reduction rates of iron
ores in a hydrogen-plasma apparatus using a graphite electrode.
At the early stages of their experiments, reduction was mainly
achieved through carbon-containing agents stemming from the
graphite electrode. Later, hydrogen acted as the major reductant.
The reduction kinetics was determined via chemical analysis of the
output gasses (mass spectrometry) and the degree of metallization
of the product reached values of approximately 70% (the remain-
ing 30% was composed of slag). The reduction kinetics of iron ores
using hydrogen plasma was also studied by Kamiya et al. [28]. Us-
ing a constant flow of an argon-hydrogen mixture (15% H2 ) in a
tungsten-equipped reactor, these authors found a reduction rate of
1.2 g/min. Our current experiments show that the iron conversion
rates for the first set of experiments (using 9 g hematite pieces)
varied from 8%wt./min (at 3.5 min) to 1.2%wt./min (at 30 min),
Fig. 10(b). Considering the input mass of 9 g, the corresponding
rates of mass loss (0.72 and 1.08 g/min) are slightly lower than
those reported in Ref. [28]. However, we had also found that the
conversion rates were substantially enhanced when using hematite
samples of higher mass (15 g). In this case, the maximum rate of
12 wt.%/min is translated into 1.8 g/min, a value which is 1.5 times
higher than that found in Ref. [28]. Hence, our results show con-
version rate values of a magnitude that has been reported also be-
fore for hydrogen plasma-based reduction of iron oxides. We also
observe that faster reduction kinetics is achievable when the input
material mass is properly adjusted in relation to the power supply
provided by the plasma arc [53]. This finding suggests to devote
more systematic studies to the influence of the ratio between ore
mass, hydrogen partial pressure and plasma parameters in terms
of voltage, power and current density.
4.3. Hydrogen direct solid-state reduction versus hydrogen plasma
reduction of molten iron ore
The reduction kinetics of iron ores using different reductant
agents (H2 , CO, solid carbon, Fe-C melt) was evaluated by Na-
gasaka et al. [29] and several other authors [54–62]. These authors
13
Acta Materialia 213 (2021) 116971
reported that the reduction rates using hydrogen can be one or
two orders of magnitude faster than those of redox reactions con-
ducted with C-containing reductants. Recently, we investigated the
direct (solid state) reduction of iron ore pellets under pure H2 flow
(0.5 L/min) at 700 °C, using a thermogravimetry experiment [21].
In that work, we chose a relative low temperature compared with
those used in industrial shaft furnaces (viz. 850–10 0 0 °C) to moni-
tor the intermediate microstructural changes prior to complete re-
duction and associate them with the rate-limiting steps of this pro-
cess. Solid-state direct reduction of iron ore pellets using molecular
hydrogen at higher temperatures has been reported in [15,18,55].
Turkdogan and Vinters [15] investigated the direct reduction of ~
15 mm-pellets at 10 0 0 °C under several gas flow rates. Bonalde
et al. [18] evaluated the reduction reaction at 850 °C using differ-
ent gasses, including pure molecular hydrogen. Results reported in
[15,18] thus provide insights into the reduction kinetics close to
those targeted in the industrial practice. The reduction kinetics re-
ported in these preceding works are reproduced as reference data
in Fig. 13(a) in comparison with the iron weight fractions obtained
in this work via hydrogen plasma-based reduction. In this figure, a
logarithmic abscissa is used to facilitate visualization and compar-
ison of the data.
Fig. 13(a) reveals that a total reduction via the direct reduction
process is achieved after 100 min of exposure to molecular hydro-
gen at 700 °C. From our preceding work, we found that the direct
reduction occurs preferentially at the external layers of the pellet
due to the short diffusion distances for O from the interface reac-
tion to the surface. Thus, at 33% reduction, iron is predominantly
formed at the shell of the pellet whereas the core remains pre-
dominantly composed of wüstite. The further 67% of reduction is
essentially given by the Fex O + H2 → x Fe + H2 O reaction and
displays notably sluggish kinetics, as reproduced in Fig. 13(b). This
is particularly due to the slow nucleation of the iron crystals (ther-
modynamic barrier) and to the sluggish mass transport, particu-
larly of the outbound transport of oxygen, throughout the com-
pact iron layer that forms as a dense shell on the wüstite (i.e.
as outer layers of the pellets) [21]. Thus, microstructure features
such as pore tortuosity, formation of dislocations, internal inter-
faces and free surfaces due to volume mismatch among the phases,
and nano-dispersed transient metal oxides were found to exert an
important role in the overall reduction kinetics by providing ad-
ditional free paths for reductant and product transport and fresh
surfaces for reaction.
Fig. 13(a) also shows that complete reduction occurs after
15 min of direct exposure to hydrogen at 850 °C (H2 flow of
2 L/min) [18]. A similar trend is observed when the process is con-
ducted at 10 0 0 °C with a H2 flow rate of 1 L/min [15]. The re-
duction kinetics can be further enhanced at 10 0 0 °C by increas-
ing the gas flow to 10 L/min so that the total conversion into iron
is obtained within 10 min [15]. This behavior is most likely due
to the faster diffusion of oxygen to the reaction interface at 850
and 10 0 0 °C rather than that at 700 °C. It should be also noted
that the magnitude of the H2 flow rates used in Refs. [15,18] is
at least twice as high as the one used in our preceding work
[21] (0.5 L/min at 700 °C, brown line in Fig. 13a), providing high
mass transfer rates of reactants to the reaction interface. In addi-
tion, Turkdogan and Vinters [15] also observed that the pore struc-
ture of the outer iron layer formed at 10 0 0 °C is coarse and not so
fine and highly dispersed as the ones observed in pellets reduced
at lower temperatures (e.g. 700 °C). The coarse pore structure of
iron facilitates gas diffusion thorough it, thus helping to overcome
the rate-limiting barrier of the process as well.
Fig. 13(b) shows the corresponding reduction rates for both, the
hydrogen molecular-based solid state reduction and the hydrogen
plasma-based reduction, where the oxide gets melted. The data re-
veal that the reduction rates through hydrogen plasma exposure
I. R. Souza Filho, Y. Ma, M. Kulse et al. Acta Materialia 213 (2021) 116971

Fig. 12. Phase transformation kinetics during hydrogen plasma-based reduction of the 15 g-hematite pieces. (a) Changes in the phase weight fractions of magnetite, wüstite,
and ferrite (b) Iron conversion kinetics and corresponding transformation rates.

Fig. 13. (a) Iron conversion kinetics during hydrogen plasma-based reduction of the 9 and 15 g-hematite pieces conducted in this study, represented by the yellow and
green curves, respectively. These plasma synthesis results are compared to the direct reduction kinetics of a hematite pellet via solid-state direct reduction using molecular
hydrogen at 700 °C (brown curve) reported in a preceding study [21]; direct reduction kinetics using molecular hydrogen at 10 0 0 °C with H2 flow rates of 10 L/min (black
line) and 1 L/min (red line) from Ref. [15], and the direct reduction kinetics of iron ore pellets with molecular hydrogen at 850 °C (2 L/min) from Ref. [18]. (b) Comparison
of the corresponding reduction rates vs. the reduction wt. fractions.

are comparable with those observed through conventional hydro-
gen direct solid-state reduction conducted at temperatures typi-
cally employed in shaft furnaces (850–10 0 0 °C) and under mod-
erate H2 flow rates. It is worth to note though that the kinetic bar-
riers associated with direct reduction (solid-state nucleation and
solid-state mass transport) do not exist in plasma reduction, as in
the latter process all reactions occur in the liquid state. Yet, hy-
drogen plasma species substantially reduce the values of G for
wüstite reduction (Section 4.1). The high reactivity of the radi-
cals and the high energy content of the hydrogen plasma is trans-
ferred to the reaction interface (molten pool), enabling local heat
release and self-supplying energy for the endothermic Fex O reduc-

14
I. R. Souza Filho, Y. Ma, M. Kulse et al.
tion [36,37], which means that the wüstite reduction using hydro-
gen plasma is exothermic and not endothermic as the solid-state
direct reduction. Thus, the exothermic net energy balance of the
hydrogen plasma reduction ensures higher efficiency in power con-
sumption when compared with that in the endothermic solid-state
direct reduction, which requires additional external energy input
to maintain the progressing reaction. All these factors explain why
the reduction process via hydrogen plasma proceeds similarly to
that of molecular hydrogen based direct reduction and reveal their
advantages in terms of energy consumption.
4.4. Micro- and nanoscale chemical evolution
In the partially reduced samples, the gangue elements are
mostly partitioned to the untransformed oxide when probed at
the micro- and nanoscale, as revealed by the EDS and APT mea-
surements, respectively. However, the chemical composition analy-
sis (Fig. 8) reveals that such gangue elements are gradually elim-
inated from the samples over the course of the reduction, i.e. the
plasma reduction process cleans the material, an advantage that is
very important for instance when reducing high Si, P and S con-
taining ores. The effect is mainly due to the relatively high vapor
pressures of most of the gangue elements that occur in iron ores,
relative to that of iron, revealing yet another thermodynamic ad-
vantage of the plasma over the conventional direct reduction pro-
cess [31,32,63]. From Fig. 6 we can learn that the last 15 min are
essential in the overall plasma-based reduction process for clean-
ing the formed iron from its retained oxide particles. The final iron
product is nearly free of gangue residues, especially the harmful
elements P and S have been removed (Figs. 6 and 8). These results
confirm preceding observations that dephosphorization and deox-
idation occur over plasma-reduction mainly through evaporation,
thus reducing the need for expensive and time-consuming sec-
ondary metallurgy operations for removing these elements [32,64].
Silicon is found in the interdendritic fayalite, at the wüstite/α -
iron phase interfaces and within the oxide particles pinned inside
bcc-iron. Interdendritic fayalite forms in the final stages of solidi-
fication, as suggested by the corresponding thermodynamics sim-
ulations (Fig. 9) and as visualized in Section 3. Si-enrichment at
the hetero-interfaces can also occur during solidification, when the
solidification front of the wüstite encounters the solid iron. This
finding can be attributed to the high diffusivity of silicon, which
leads to its partitioning to the ever-decreasing fractions of liquid.
Although silicon is quite abundant in the microstructure, its con-
tent also decreases during the course of reduction (Fig. 8). In this
context, the reduction of the anionic complex SiO4 −4 into metal-
lic silicon is not promoted and it is eliminated from the samples
through the gaseous SiO suboxide [32,64].
5. Summary and conclusions
We investigated the chemistry and phase evolution during hy-
drogen plasma-based reduction of commercial hematite. The fol-
lowing conclusions are drawn:
• The iron conversion kinetics is dependent on the balance be-
tween the initial hematite mass and the arc power. The reduc-
tion kinetics of the 9 g hematite specimens is relative sluggish
during the final 25% of reduction. By increasing the weight of
the initial hematite pieces to 15 g, the reduction kinetics is
substantially increased, occurring two times faster. The max-
imum conversion rate was 12 wt.%/min, which is translated
into 1.8 g/min. Total reduction to pure iron was achieved af-
ter 15 min. It appears that larger molten baths facilitate trans-
port of oxygen to the reaction interface. For both sets of exper-
iments, using 9 g and 15 g-pieces, wüstite reduction is the rate
limiting thermodynamic barrier of the process.
15
Acta Materialia 213 (2021) 116971
• The reduction rate of hydrogen plasma-based reduction, using
only a small H2 partial pressure of 10%, is comparable with
those ones observed in solid-state direct reduction of hematite
conducted at temperatures currently employed in shaft fur-
naces (850–10 0 0 °C). The rate-limiting barriers (i.e. diffusion of
oxygen species in the solid state) in the direct reduction pro-
cess do not exist in the hydrogen plasma process. In the lat-
ter case, the hematite is melted and reduced simultaneously.
Also, the net energy balance of iron oxide reduction via hydro-
gen plasma is exothermic and not endothermic as in the case
of the solid-sate direct reduction. This fact ensures higher effi-
ciency in power consumption of the former process.
• The gangue elements of the hematite are gradually removed
during reduction, due to their high vapor pressure. Silicon en-
richment was identified in the interdendritic fayalite, at the
wüstite/bcc-iron hetero-interfaces and in the oxide particles in-
side iron. Since silicon oxide reduction is not readily feasible,
it is majorly evaporated in the form of SiO gas. Phosphorus and
sulfur concentrations were also remarkably reduced from 5 min
onwards and the final obtained iron is nearly free of gangue el-
ements.
Our observations shed new light on the underlying thermody-
namics and kinetics of hydrogen plasma-based reduction of iron
ores, providing new insights into the feasibility of such approach
as a carbon-neutral alternative for green iron production.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgements
We thank Monika Nellessen and Katja Angenendt for their sup-
port to the metallography lab and SEM facilities at MPIE. We are
grateful to Uwe Tezins, Christian Broß, and Andreas Sturm for
their support to the FIB & APT facilities at MPIE. We are grate-
ful to Benjamin Breitbach for the support to the X-ray diffrac-
tion facilities at MPIE. BG acknowledges financial support from the
ERC–CoG-SHINE-771602. HS acknowledges finacial support through
the Heisenberg Programm of the Deutsche Forschungsgemein-
schaft.
Supplementary materials
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.actamat.2021.116971.
References
[1] E. Basson, World Steel in Figures, Yearb. World Steel Assoc. 2020 (2020) 1–8.
[2] D. Raabe, C.C. Tasan, E.A. Olivetti, Strategies for improving the sustainability of
structural metals, Nature 575 (2019) 64–74, doi:10.1038/s41586- 019- 1702- 5.
[3] M. Ashby, Mater. Environ. (2013) Seond Edition, doi:10.1016/C2010- 0- 66554- 0.
[4] M. Fischedick, J. Marzinkowski, P. Winzer, M. Weigel, Techno-economic eval-
uation of innovative steel production technologies, J. Clean. Prod. 84 (2014)
563–580, doi:10.1016/j.jclepro.2014.05.063.
[5] V. Vogl, M. Åhman, L.J. Nilsson, Assessment of hydrogen direct reduction
for fossil-free steelmaking, J. Clean. Prod. 203 (2018) 736–745, doi:10.1016/j.
jclepro.2018.08.279.
[6] E. Commissie, A Roadmap for Moving to a Competitive Low Carbon Economy
in 2050, Europese Commissie, Brussel, 2011.
[7] C. Hoffmann, M. van Hoey, B. Zeumer, Decarbonization challenge for steel Hy-
drogen as a solution in Europe, McKinsey & Company (2020).
[8] F. Patisson, O. Mirgaux, Hydrogen ironmaking: how it works, Metals (Basel) 10
(2020) 1–15, doi:10.3390/met10070922.
[9] A. Bhaskar, M. Assadi, H.N. Somehsaraei, Decarbonization of the iron and steel
industry with direct reduction of iron ore with green hydrogen, Energies. 13
(2020). https://doi.org/10.3390/en13030758.
I. R. Souza Filho, Y. Ma, M. Kulse et al.
[10] D. Spreitzer, J. Schenk, Reduction of iron oxides with hydrogen—a review, Steel
Res. Int. 90 (2019) 1900108, doi:10.10 02/srin.20190 0108.
[11] A.A. El-Geassy, V. Rajakumar, Influence of particle size on the
gaseous reduction of wustite at 90 0-110 0.DEG.C., transactions
of the Iron and Steel Institute of Japan. 25 (1985) 1202–1211.
https://doi.org/10.2355/isijinternational1966.25.1202.
[12] S.K. El-Rahaiby, Y.K. Rao, The kinetics of reduction of iron oxides at moderate
temperatures, Metall. Trans. B 10 (1979) 257–269, doi:10.1007/BF02652470.
[13] A.A. El-Geassy, M.I. Nasr, Influence of the original structure on the kinetics
of hydrogen reduction of hematite compacts, Trans. Iron Steel Inst. Japan 28
(1988) 650–658, doi:10.2355/isijinternational1966.28.650.
[14] J.O. Edstrom, The mechanism of reduction of iron oxides, J. Iron Steel Inst. 175
(1953) 289.
[15] E.T. Turkdogan, J.V. Vinters, Gaseous reduction of iron oxides: part I. reduction
of hematite in hydrogen, Metallurg. Mater. Trans. B 2 (1971) 3175–3188, doi:10.
1007/BF02814970.
[16] E.T. Turkdogan, J.v. Vinters, Gaseous reduction of iron oxides: part III.
reduction-oxidation of porous and dense iron oxides and iron, Metall. Trans.
3 (1972) 1561–1574, doi:10.1007/BF02643047.
[17] M. Bahgat, M.H. Khedr, Reduction kinetics, magnetic behavior and morpholog-
ical changes during reduction of magnetite single crystal, Mater. Sci. Eng.: B
138 (2007) 251–258, doi:10.1016/j.mseb.2007.01.029.
[18] A. Bonalde, A. Hnriquez, M. Manrique, Kinetic analysis of the iron oxide reduc-
tion using hydrogen-carbon monoxide mixtures as reducing agent, ISIJ Int. 45
(2005) 1255–1260, doi:10.2355/isijinternational.45.1255.
[19] N.G. Gallegos, M.A. Apecetche, Kinetic study of haematite reduction by hydro-
gen, J. Mater. Sci. 23 (1988) 451–458, doi:10.1007/BF01174669.
[20] B. Hou, H. Zhang, H. Li, Q. Zhu, Study on Kinetics of Iron Oxide Reduction
by Hydrogen, Chin. J. Chem. Eng. 20 (2012) 10–17, doi:10.1016/S1004-9541(12)
60357-7.
[21] S.-.H. Kim, X. Zhang, Y. Ma, I.R. Souza Filho, K. Schweinar, K. Angenendt, D. Vo-
gel, L.T. Stephenson, Ayman A. El-Zoka, J.R. Mianroodi, M. Rohwerder, B. Gault,
D. Raabe, Influence of microstructure and atomic-scale chemistry on the direct
reduction of iron ore with hydrogen at 700°C, Acta Mater. 212 (2021) 116933,
doi:10.1016/j.actamat.2021.116933.
[22] J. Hnderson, Hydrogen reduction characteristics of natural and artificial hæ-
matites, Nature (1961) 191, doi:10.1038/191268a0.
[23] P. Rajput, B. Bhoi, S. Sahoo, R.K. Paramguru, B.K. Mishra, Preliminary investiga-
tion into direct reduction of iron in low temperature hydrogen plasma, Iron-
mak. Steelmak. 40 (2013) 61–68, doi:10.1179/1743281212Y.0 0 0 0 0 0 0 023.
[24] P. Rajput, K.C. Sabat, R.K. Paramguru, B. Bhoi, B.K. Mishra, Direct reduction of
iron in low temperature hydrogen plasma, Ironmak. Steelmak. 41 (2014) 721–
731, doi:10.1179/1743281214Y.0 0 0 0 0 0 0186.
[25] K.C. Sabat, P. Rajput, R.K. Paramguru, B. Bhoi, B.K. Mishra, Reduction of oxide
minerals by hydrogen plasma: an overview, plasma chemistry and plasma pro-
cessing. 34 (2014) 1–23. https://doi.org/10.1007/s11090-013-9484-2.
[26] K. Hassouni, A. Gicquel, M. Capitelli, J. Loureiro, Chemical kinetics and
energy transfer in moderate pressure H 2 plasmas used in diamond
MPACVD processes, Plasma Sources Science and Technology. 8 (1999) 494–512.
https://doi.org/10.1088/0963-0252/8/3/320.
[27] K. Badr, in: PhD Dissertation, University of Leoben, Austria, 2007, pp. 1–169.
[28] K. Kamiya, N. Kitahara, I. Morinaka, K. Sakuraya, M. Ozawa, M. Tanaka, Reduc-
tion of Molten Iron Oxide and FeO Bearing Slags by H2-Ar Plasma, Trans. Iron
Steel Inst. Japan 24 (1984) 7–16, doi:10.2355/isijinternational1966.24.7.
[29] T. Nagasaka, M. Hino, S. Ban-Ya, Interfacial kinetics of hydrogen with liquid slag
containing iron oxide, Metallurg. Mater. Trans. B 31 (20 0 0) 945–955, doi:10.
10 07/s11663-0 0 0-0 071-6.
[30] M.N. Seftejani, J. Schenk, Kinetics of molten iron oxides reduction using hydro-
gen, La Metallurgia Italiana (2018) 5–14.
[31] J.F. Plaul, W. Krieger, E. Bäck, Reduction of fine ores in argon-hydrogen plasma,
Steel Res. Int. 76 (2005) 548–554, doi:10.1002/srin.200506055.
[32] H. Hiebler, J.F. Plaul, Hydrogen plasma smelting reduction - an option for steel-
making in the future, Metalurgija 43 (2004) 155–162.
[33] K.U. Maske, J.J. Moore, The application of plasmas to high temperature reduc-
tion metallurgy, High Temp. Technol. 1 (1982) 51–63, doi:10.1080/02619180.
1982.11753180.
[34] M. Naseri Seftejani, J. Schenk, Thermodynamic of liquid iron ore
reduction by hydrogen thermal plasma, metals. 8 (2018) 1051.
https://doi.org/10.3390/met8121051.
[35] M. Naseri Seftejani, J. Schenk, M.A. Zarl, Reduction of haematite using
hydrogen thermal plasma, Materials (Basel) 12 (2019) 1608, doi:10.3390/
ma12101608.
[36] K.C. Sabat, A.B. Murphy, Hydrogen plasma processing of iron ore, Metallurg.
Mater. Trans. B 48 (2017) 1561–1594, doi:10.1007/s11663- 017- 0957- 1.
[37] Y. Zhang, W. Ding, S. Guo, K. Xu, Reduction of metal oxide in nonequilibrium
hydrogen plasma, China Nonferrous Metal 14 (2004) 317–321.
[38] P.R. Behera, B. Bhoi, R.K. Paramguru, P.S. Mukherjee, B.K. Mishra, Hydrogen
Plasma Smelting Reduction of Fe2 O3 , Metallurg. Mater. Trans. B 50 (2019) 262–
270, doi:10.1007/s11663- 018- 1464- 8.
16
Acta Materialia 213 (2021) 116971
[39] M. Naseri Seftejani, J. Schenk, D. Spreitzer, M.Andreas Zarl, Slag formation dur-
ing reduction of iron oxide using hydrogen plasma smelting reduction, Mate-
rials (Basel) 13 (2020) 935, doi:10.3390/ma13040935.
[40] M.P. Moody, B. Gault, L.T. Stephenson, R.K.W. Marceau, R.C. Powles, A.V.
Ceguerra, A.J. Breen, S.P. Ringer, Lattice rectification in atom probe tomogra-
phy: toward true three-dimensional atomic microscopy, microscopy and mi-
croanalysis. 17 (2011) 226–239. https://doi.org/10.1017/S1431927610094535.
[41] A.V. Ceguerra, A.C. Day, S.P. Ringer, Assessing the spatial accuracy of the
reconstruction in atom probe tomography and a new calibratable adap-
tive reconstruction, Microsc. Microanal. 25 (2019) 309–319, doi:10.1017/
S14319276190 0 0369.
[42] M. Bachhav, R. Danoix, F. Danoix, B. Hannoyer, S. Ogale, F. Vurpillot, Inves-
tigation of wüstite (Fe1−x O) by femtosecond laser assisted atom probe to-
mography, Ultramicroscopy 111 (2011) 584–588, doi:10.1016/j.ultramic.2010.11.
023.
[43] M. Bachhav, F. Danoix, B. Hannoyer, J.M. Bassat, R. Danoix, Investigation of O-
18 enriched hematite (α -Fe2O3) by laser assisted atom probe tomography, Int.
J. Mass Spectrom. 335 (2013) 57–60, doi:10.1016/j.ijms.2012.10.012.
[44] B. Gault, D.W. Saxey, M.W. Ashton, S.B. Sinnott, A.N. Chiaramonti, M.P. Moody,
D.K. Schreiber, Behavior of molecules and molecular ions near a field emitter,
New J. Phys. 18 (2016) 033031, doi:10.1088/1367-2630/18/3/033031.
[45] S. Hayashi, Y. Iguchi, Influence of gangue species on hydrogen reduction rate
of liquid wustite in gas-conveyed systems, ISIJ Int. 35 (1995), doi:10.2355/
isijinternational.35.242.
[46] M.E. Choi, H.Y. Sohn, Development of green suspension ironmaking
technology based on hydrogen reduction of iron oxide concentrate:
rate measurements, Ironmak. Steelmak. 37 (2010) 81–88, doi:10.1179/
030192309X12506804200663.
[47] H. Nishiyama, T. Sawada, H. Takana, M. Tanaka, M. Ushio, Computational sim-
ulation of arc melting process with complex interactions, ISIJ Int. 46 (2006)
705–711.
[48] J.-.P. Bellot, L. Décultot, A. Jardy, S. Hans, E. Doridot, J. Delfosse, N. McDon-
ald, Numerical Simulation of the Plasma Arc Melting Cold Hearth Refining Pro-
cess (PAMCHR), Metallurg. Mater. Trans. B 51 (2020) 1329–1338, doi:10.1007/
s11663- 020- 01866- 0.
[49] M. Tanaka, M. Ushio, J.J. Lowke, Numerical study of gas tungsten arc plasma
with anode melting, Vacuum 73 (2004) 381–389, doi:10.1016/j.vacuum.2003.
12.058.
[50] S.C. Chu, S.S. Lian, Numerical analysis of temperature distribution of plasma
arc with molten pool in plasma arc melting, Comput. Mater. Sci. 30 (2004)
441–447, doi:10.1016/j.commatsci.2004.03.014.
[51] R.M. Urusov, T.E. Urusova, Computation of steady flow in the molten pool at
electric-arc heating, Thermophys. Aeromech. 14 (2007) 265–276.
[52] A.B. Murphy, M. Tanaka, K. Yamamoto, S. Tashiro, T. Sato, J.J. Lowke, Modelling
of thermal plasmas for arc welding: the role of the shielding gas properties
and of metal vapour, J. Phys. D Appl. Phys. 42 (2009) 194006, doi:10.1088/
0022-3727/42/19/194006.
[53] N.A. Barcza, T.R. Curr, R.T. Jones, Metallurgy of open-bath plasma processes,
Pure Appl. Chem. 62 (1990) 1761–1772, doi:10.1351/pac199062091761.
[54] T. Nagasaka, Y. Iguchi, S. Ban-Ya, Rate of reduction of liquid wüstite with CO,
Tetsu-to-Hagane 71 (1985) 204–211, doi:10.2355/tetsutohagane1955.71.2_204.
[55] S. Ban-Ya, Y. Iguchi, T. Nagasaka, Rate of reduction of liquid wüstite with hy-
drogen, Tetsu-to-Hagane 70 (1984) 1689–1696, doi:10.2355/tetsutohagane1955.
70.14_1689.
[56] J.-.C. Lee, D.-.J. Min, S.-.S. Kim, Reaction mechanism on the smelting reduction
of iron ore by solid carbon, Metallurg. Mater. Trans. B 28 (1997) 1019–1028,
doi:10.1007/s11663- 997- 0056- 9.
[57] T. Soma, Smelting reduction of iron ore, Bull. Jpn. Inst. Metals 21 (1982)
620–625.
[58] F. Tsukihashi, M. Amatatsu, T. Soma, Reduction of molten iron ore with carbon,
Tetsu-to-Hagane 68 (1982) 1880–1888, doi:10.2355/tetsutohagane1955.68.14_
1880.
[59] Y. Sasaki, T. Soma, The kinetics of the reduction of FeO in CaO-SiO2
slag with solid carbon, Tetsu-to-Hagane 64 (1978) 376–384, doi:10.2355/
tetsutohagane1955.64.3_376.
[60] K. Takahashi, M. Amatatsu, T. Soma, Reduction of Molten Iron Ore
with Solid Carbon, Tetsu-to-Hagane 61 (1975) 2525–2530, doi:10.2355/
tetsutohagane1955.61.11_2525.
[61] T. Nagasaka, Y. Iguchi, S. Ban-Ya, Effect of additives on the rate of reduction
of liquid iron oxide with CO, Tetsu-to-Hagane 75 (1989) 74–81, doi:10.2355/
tetsutohagane1955.75.1_74.
[62] A. Sato, G. Aragane, K. Kamihira, S. Yoshimatsu, Reduction rate of molten iron
oxide by the solid carbon or the carbon in molten iron, Tetsu-to-Hagane 73
(1987) 812–819, doi:10.2355/tetsutohagane1955.73.7_812.
[63] Y. Nakamura, M. Ito, H. Ishikawa, Reduction and dephosphorization of molten
iron oxide with hydrogen-argon plasma, Plasma Chem. Plasma Process. 1
(1981) 149–160.
[64] F.K. McTaggart, Reduction of silica in a hydrogen discharge, Nature 201 (1964)
1320–1321, doi:10.1038/2011320a0.